As you’ll no doubt have inferred from the post immediately preceding this one, I’ve been thinking about Vannevar Bush again.
I doubt there’s been a worthwhile work on the history of computing in the history of computing that did not cite his “As We May Think,” the July 1945 Atlantic article in which he introduced the world to the “memex.” Its insights were that far ahead of their time, that foundational to so much that would come afterward.
Just what did Bush say in “As We May Think”? It’s one of those texts – like Don Norman’s Design of Everyday Things, like (yes, again) The Death and Life of Great American Cities – so seminal that you primarily remember its emotional impact, and not always its argument. In the course of writing once again a book in which I’ll wind up citing him, I thought it’d be more than worthwhile to go back to the source and pin down just what it is that I’m sending people to see.
Bush started by describing a problem we now recognize as essentially one of knowledge management: by the mid-twentieth century, there was already too much specialized information in the world for one person to keep fully abreast of developments in their own field, let alone communicate meaningfully across disciplines. (Forget talking to nonspecialists entirely.) As a result, the collective body of knowledge we call “science” lagged badly: we simply did not know what it is that we already knew. This is, still, something most of us can attest to; I felt it acutely the moment I began to cobble together the arguments that are supposed to wind up as a book.
Where Bush differed from most people who have ever complained about this fact of life, though, was in daring to sketch out a proposed solution. His first step was to establish the contemporary existence of certain technologies crucial to the new thing about to be proposed – in a sense, the small pieces he’s about to loosely connect. A few rapid-fire paragraphs suffice to acquaint the reader with the input and output modalities he will rely on: voders and vocoders, short focal-length lenses, photocells and thermionic tubes.
In this, he wasn’t interested in elegance, economy or size, at least not initially; this step was just an existence proof. What was necessary to demonstrate was simply that the sort of information-processing operations he had in mind could (in principle, anyway) be accomplished with extant techniques; optimization could wait.
As long as the methods to do what he wished existed in the world, Bush was certain that they would later be improved. And while he may not yet have had an explicitly-enunciated Moore’s Law to back him up, there’s no doubt that he felt information technology was already embarked on a curve of continual improvement: “[i]t would be a brave man who could predict that such [technologies] will always remain clumsy, slow, and faulty…”
But he does manage to establish his kit of parts, and this buys him a fully articulated information-processing ecology – albeit one that doesn’t quite go as far as it might. (Hey, what do you want for a single article?) If the telegraph was “the Victorian Internet,” what can be glimpsed here is a fedora-sporting, Truman-era Internet of solenoids, punchcards and radio waves.
It would have been something to see. This much is clear from the two use cases he offers us: first a scientific researcher equipped with mobile image-capture and annotation, then a point-of-purchase transaction, taking place in a great department store. The latter scene is particularly poignant, in that he specifies a system that provides for inventory management, sales commissions, accounting and billing…and uses it to print out a regular monthly bill to be sent to the customer through postal mail! (If Bush here seizes up with a thump against the limits of his worldview, it’s beyond dispute that he still managed to limn an entire computational ecology whose real-world equivalents wouldn’t be deployed until the early 1970s.)
As radical as these ideas must have seemed to the audience of 1945, they’re just appetizers. The main course would be both the crowning achievement of Vannevar Bush’s career and the reason why “As We May Think” is still read and remembered: a device he called the memex. This would be “a sort of mechanized private file and library…an enlarged intimate supplement to memory.” (Maybe we’d now call it a PAA, a “personal analogue assistant.”)
The memex, as Bush described it, was an ordinary office desk augmented with an internal, microminiaturized archive (of Library of Congress extent) and tilted display and input screens. It could “presumably” be operated remotely. It allowed not merely the review but also the annotation of whatever body of knowledge had been stored in it. And most critically, via a clever indexing scheme, it allowed one fact to be associated with another. And this was how Bush proposed to deal with the urgent problem of information overload with which he’d begun the article.
The insuperable problem of placing data in storage was that it could only be in one physical place at a time. Librarians, certainly, had over the course of centuries evolved practices aimed at classifying and arranging books such that they could be stored and their locations reliably reacquired with a relative minimum of effort. But such classifications, and the physical nature of knowledge as manifest in books, meant that a history (say) of sexuality in Eastern religions could appear either under “Human Sexuality” or “Religion,” but not under both – not unless you could afford to have multiple copies. There was just the one official and unchanging pathway: at mid-century library science had not yet fully wrestled with the idea of multiplicity, with the idea that one text could present multiple facets of meaning, or acquire multiple meanings that varied by reader and context.
By contrast, decades before hypertext, HTML or the World Wide Web, Bush’s memex proposed to allow its user to organize bodies of knowledge along individualized “trails,” each permanent and immediately available for lookup in exactly the same way that we might use a Web browser’s bookmarks. Moreover, any one item could belong to an arbitrary number of different contextual trails – trails that could be stored side-by-side in one’s own memex, or passed entire to friends and colleagues for them to explore at their leisure, like an analogue del.icio.us.
(One might also argue that, in his provision for a very natural way of organizing knowledge – a schema for storing and recalling information that worked with what was known about human cognition, rather than fighting upstream against it – Bush had unwittingly invented the field of user experience, but we’ll leave that battle for a different day.)
However visionary Bush was in other respects, he was undeniably a man of his time. Except in a vague and speculative coda, he completely failed to anticipate digital technology. All of the innovations he depends on as componentry of the memex – forehead-mounted cameras, microfilm Brittanicas – are almost comically dependent on the tropes and methods of the analogue world. And, though we can hardly blame him for it, he remained mired in the gender politics of his day: Bush’s future world is one in which investigators are always “he” and the drudgework of data entry is given over to “roomfuls of girls.”
Nevertheless, he did manage to leave us the memex, in which both hypertext and a whole approach to the human use of informatics begin. And he also gave us something else: although “As We May Think” is generally cited as the origin of hypertext, it also depicts a mobile investigator roaming the world freely, “uploading” information to a network there to catch it. Finally, all but buried in his conclusion, he offers up the critical insight on which a man named Mark Weiser would build the doctrine of computational ubiquity some forty years downstream: work with information-processing devices is both more effective and more enjoyable if the user “can reacquire the privilege of forgetting the manifold things he does not need to have immediately at hand, with some assurance that he can find them again if they prove important.”
It’s for both this and his frank description of the memex as outboard memory augmentation, rather than the relatively prosaic idea of integrating knowledge-processing devices into a common desk, that I think of Vannevar Bush as belonging properly to the history of ubicomp. Now to draw the lines connecting the memex and the ubicomp that grew out of it to place-making and the city.